Research On The Identification Of Sui Language

Guizhou is a multi-ethnic province.Most ethnic minorities have their own language.The language of ethnic minorities is not only a valuable cultural heritage,but also a tool for cultural exchange of the ethnic group.However,with the acceleration of China’s urbanization process,many ethnic minority populations in remote rural areas have moved from rural areas to urban work and life,and daily communication has been gradually replaced by Mandarin,especially for the next generation.The inheritance and protection of ethnic languages is facing a huge test.This article uses signal processing,pattern recognition and speech processing technology to study the key pronunciation of Guizhou aquatic language(abbreviated as Sui language),which lays a good foundation for the next step of automatic recognition of aquatic language.The main research work of the thesis is as follows:(1)Establish a Sui language speech corpus.Use recorded voice software(this article uses Feifeng software)to make multiple recordings of Sui language phrases or phrases to ensure the validity of subsequent test results and the accuracy of analysis.After the recording is completed,the recorded voice signal is annotated using praat software,and the initials,finals,and key values of the Sui language included in each voice are marked out to form a Sui language speech corpus.(2)A method for preprocessing of speech information in Sui language is proposed.First use the double threshold method to detect Voice Activity Detection of the speech,and extract the sound wave with the Sui language speech information.Then the sampling,quantization,pre-emphasis,framing and windowing processing of the speech signal after passing through Voice Activity Detection.(3)Through the analysis of the Sui language speech signal,the seven main vowel features in the Sui language speech are extracted,and the 24-dimensional feature vectors of these vowels are obtained by using the MFCC method.(4)The Hidden Markov Model(HMM)is used to establish a classifier for recognition of Sui language vowels.In this paper,1120 male and female hydrophone speech data are selected for experiment.The 24-dimensional vector extracted by MFCC is used as the training set of the HMM algorithm model to establish a hydrophone vowel recognition classifier to automatically classify and recognize the seven main hydrophone speech voices.The experimental results show that the established method has a better recognition effect on the vowels of the Sui language.